Among older Americans, adverse drug events (ADEs) represent a significant public health challenge. Estimates indicate that ADEs may occur in up to 35% of older outpatients [1
] and cause or contribute to 6%–12% of this cohort’s hospital admissions [2
]. Predictably, ADEs carry economic consequences. Medication-related morbidity and mortality may be responsible for 16% of the United States’ annual healthcare costs, representing a $
500 billion annual toll [3
]. Fortunately, the US healthcare system should be able to mitigate the economic and clinical consequences of ADEs. Consistent data suggest that the majority of ADEs are both predictable and preventable [1
The predictable and preventable nature of ADEs is a call to action. A risk stratification tool that identifies individuals at risk could help policy makers, healthcare organizations, and payers curb the negative clinical and economic consequences of ADEs. While ADE risk prediction tools do exist, a disproportionate amount have been validated in hospital settings and, as recently reported, several of them cannot accurately predict the risk of drug-related problems [2
]. Moreover, many of the existing tools utilize chronic comorbidities or laboratory values, such as renal failure, heart failure, or white blood cell count, to predict ADEs [5
]. While these variables may accurately predict ADE risk, they are not necessarily modifiable or easily actionable. Thus, they are of dubious clinical utility. By contrast, the MedWise Risk Score™ (MRS) [9
], described by Cicali et al., uses potentially alterable pharmacological characteristics of patients’ medications for risk stratification, making it potentially easier for healthcare providers to mitigate medication risk. However, until now, associations with ADEs and other important outcomes (e.g., medical expenditures, emergency department visits, hospitalizations) have not been reported in controlled studies.
The US-government funded Program of All-Inclusive Care for the Elderly (PACE) is an ideal setting to examine the MRS. PACE provides supportive services to community-dwelling older adults who are certified by the state in which they reside to require a “nursing home level of care” [10
]. Given a capitated payment model, PACE organizations aim to avoid unnecessary medical expenditures, nursing home institutionalization, and frequent hospital visits [10
]. PACE organizations are at great risk of unnecessary spending associated with medication-related morbidity given their participants’ significant comorbidities [11
] and prevalence of medication-related problems [12
]. A tool like the MRS could identify these organizations’ high-risk members and provide guidance to mitigate risk. Therefore, our objective here is to examine whether the MRS correlates with ADEs as well as the outcomes that are important to PACE and other government-funded healthcare programs: total cost, emergency department (ED) visits, hospitalizations, and hospital length of stay.
During 2018, 11,988 patients were identified in CareKinesis’ total census. After making various exclusions, which are outlined in Figure 1
, 1965 PACE participants remained in the analytical sample. PACE participants included in this analysis were 65.2% (n
= 1282) female and, on average 76.8 ± 9.9 years old. Participants represented a total of 12 PACE organizations, which are geographically dispersed throughout the United States and vary in overall census size. Table 1
highlights participant demographics. Figure 2
depicts the overall distribution of the MRS among PACE participants. Overall, MRS ranged from 2 to 40 in the total sample and, on average, participants had an MRS of 18.5 ± 7.8.
During 2018, 128 (6.5%) PACE participants experienced at least one ADE. A total of 54 unique ADE-related ICD-10 codes were identified in the 2018 claims. The ICD-10 codes that described ADEs were often related to opioids, extrapyramidal symptoms, anticoagulants, skin eruptions, and psychoactive medications. These symptoms and medication classes accounted for at least 50% (n
= 85) of all (n
= 170) ADE claims. During our analysis, we found patients with ADE codes for T78.4 (allergy, unspecified), T80.2 (infections following infusion, transfusion, or therapeutic injection), and T40.1 (heroin poisoning). These codes were excluded from our analysis because such ADEs are unrelated to the MRS. The top ADEs and their corresponding ICD-10 codes can be viewed in Table 2
. All remaining ADE codes are provided in Appendix A
, Table A1
The logistic regression analysis found that the odds of having one or more ADEs over the course of the year increases by 8.6% per every point increase in the MRS (OR = 1.086, 95% CI: 1.060, 1.113; p
< 0.001). A weighted linear regression was also calculated, using MRS to predict the total number of participants with at least one ADE. A significant regression equation was found (F (1, 36) = 31.8; p
< 0.001), with an adjusted R2
of 0.454. The number of PACE participants predicted to have at least one annual ADE is equal to –0.0261 + 0.0049x, where x is the MRS. Therefore, every point increase in the MRS corresponds to an additional 4.9 participants per 1000 PACE participants with at least one ADE per year. The ADE vs. MRS weighted regression can be viewed in Figure 3
. The area under the ROC curve, which quantifies the ability of the MRS to predict the presence of ADEs, was 0.67, and is depicted in Appendix A
, Figure A1
We also assessed the MRS’ relationship to a variety of other pertinent risk outcomes, which can be viewed in Figure 4
, Figure 5
, Figure 6
and Figure 7
. Notably, a significant association was observed when MRS was regressed against total annual medical expenditures. Figure 4
depicts this weighted linear regression. A significant regression equation was found (F (1, 36) = 105.8; p
< 0.001), with an adjusted R2
of 0.739. PACE participants predicted total annual facility and physician costs as equal to 7091.50 + 1036.90x USD, where x is the MRS. Therefore, PACE participants’ annual costs increase $
1036.90 USD for every point of MRS. Figure 5
, Figure 6
and Figure 7
show all other weighted regressions calculated to predict various risk outcomes based on MRS. In summary, we observed a significant positive correlation between the MRS and various outcomes, including annual ED visits, all-cause hospital admissions, and hospital length of stay.
This retrospective study of PACE administrative medical claims demonstrates that a novel MRS derived strictly from a medication regimen’s pharmacological properties was significantly associated with ADE occurrence among medically-complex, community-dwelling older adults enrolled in a PACE program. Specifically, we found that every point increase in the MRS corresponded to nearly five additional participants having at least one ADE in the year out of every 1000 participants. However, this is likely a gross underestimation. The literature indicates that ADEs are widely underreported, with several studies citing underreporting rates that exceed 90% [39
]. To account for potential underreporting and given established associations with higher healthcare utilization and cost, we anticipated that a higher MRS would be associated with higher costs, hospital admissions, ED visits, and hospital length of stay. Our results supported this idea, as each point increase in the MRS corresponded to over $
1000 USD in additional annual medical spending, three additional annual ED visits per 100 participants per year, and two additional hospitalizations per 100 participants per year. Collectively, these results indicate that it is possible to utilize the pharmacological properties of a drug regimen to risk stratify PACE participants and predict a host of important and relevant outcomes pertaining to medication-related morbidity.
The associations identified in this study have important implications for PACE organizations, which are at full financial risk for participant outcomes due to capitated payments from CMS and state Medicaid programs [10
]. Projections through 2025 estimate that, in general, costs of medical resources (e.g., hospitalizations and emergency room visits) will increase, while many states could face future annual Medicaid funding restrictions [3
]. This combination will place immense pressure on PACE to engage in cost avoidance and reduction activities [13
]. Such activities must not ignore medication-related morbidity, which could cost the healthcare system an additional $
2500 USD for each instance of non-optimized drug therapy [3
]. Given that resolution of medication-related problems may help avoid unnecessary medical expenditures [3
], deploying consistent medication risk identification and mitigation strategies in PACE is vital.
While the utilization of such strategies remains inconsistent in this setting [10
], some evidence indicates that PACE healthcare providers (HCPs) desire comprehensive and consistent support to help promote safer pharmacotherapy. Sloane et al. found that PACE physicians frequently reported difficulties in understanding how patients’ complex morbidity can alter risk of adverse health outcomes or wasted spending. This lack of understanding can lead to treatment that is inappropriate, ineffective, or unnecessary [42
]. Physicians outside of PACE have echoed these sentiments regarding the pharmaceutical care of the medically-complex older adult. Specifically, polypharmacy resulting from multi-morbidity makes medication management in this cohort especially challenging [43
]. Underscoring these challenges, a recent study found that PACE physicians and nurse practitioners frequently initiated consultation with pharmacists for information and advice when faced with medication safety uncertainties [16
]. Notably, their inquiries commonly pertained to opioids and psychoactive drugs [16
], which are medication classes responsible for the disproportionate amount of ADE claims identified in the present study.
The MRS can help solve these problems. First, PACE HCPs can now use the MRS to identify participants who are at the greatest risk of medication morbidity and healthcare utilization. Therefore, PACE organizations and pharmacists can properly allocate efforts and resources to the participants who have the greatest need for intervention. For instance, pharmacists and HCPs can focus initial medication reviews on 20% of the population at greatest risk. In this study, this would correspond to participants with an MRS ≥26. Second, point-of-care access to the MRS provides HCPs with actionable information to make geriatric medication management less challenging. The MRS is comprised of five risk factors that highlight aspects of the regimen that need the most intervention or attention. For example, if the MRS identifies that a PACE participant is at high risk due to competitive inhibition, HCPs can act to resolve any present interaction(s).
By identifying tangible ways to mitigate medication-related morbidity among the most “at risk” PACE participants, the MRS can profoundly influence HCP behavior. For example, a survey of PACE physicians found that the overwhelming majority (>85%) of respondents were more likely to deprescribe medications or reconsider medication choice as a result of having access to the MRS [46
]. When PACE pharmacists were supported with the MRS in a CDSS, HCPs accepted about 80% of pharmacists’ recommendations to resolve medication-related problems that frequently involved broad medication safety concerns (69%), such as adverse drug reactions (18%) and pharmacokinetic drug interactions (21%) [12
]. The interactions frequently involved opioids, anticoagulants, and psychoactive medications [12
], which were top contributors to ADEs in the present study. Collectively, this suggests that the MRS helps PACE HCPs and pharmacists identify and mitigate relevant medication risk.
Our analysis is not without limitations, which carry pertinent future research directives. First, retrospectively detecting ADEs with a conservative set of ICD-10 codes may have biased our results and underestimated the true ADE occurrence. As aforementioned, administrative data have a low sensitivity to detect ADEs since it is well-established that ADEs are underreported by HCPs at the point of care [38
]. Since ADEs are a difficult factor to capture through administrative claims alone, this could explain why our area under the ROC curve was fair at 0.67. To our knowledge, other risk scores only examined ADE-induced hospitalizations or inpatient-acquired ADEs, making comparisons difficult. Nevertheless, it is noteworthy that our metric was similar to the area under the ROC curve reported by Parameswaran Nair et al. (0.70), who designed a risk score to detect ADE-related hospitalizations among older outpatients whereby ADEs were clinically validated by expert consensus chart review and patient interviews [5
]. Future studies should assess the MRS in a prospective manner, with clinical ADE validation. Nevertheless, associations with ADEs were still statistically significant despite using ICD-10 codes. Therefore, the associations observed between the MRS and the other negative risk outcomes can be explained, at least in part, by medication-related morbidity. Future research will need to validate the MRS against other ADE-specific risk scores [5
] and various risk indices highly relevant to PACE (e.g., Hierarchical Condition Category scores). Finally, additional studies will be needed in alternative cohorts. Given the uniqueness of the PACE cohort, our results may not be generalizable to inpatient older adults and healthier community-dwelling older adults.